The Agent Skills Directory

[DYNAMIC_EXECUTION]: The Python script scripts/generate_cost_report.py includes a generate_script method that takes content provided via command-line arguments and embeds it into a new shell script. This allows for the runtime generation of executable code based on potentially untrusted or unvalidated input.
[PRIVILEGE_ESCALATION]: The script scripts/generate_cost_report.py programmatically modifies file permissions using file_path.chmod(0o755) to make the dynamically generated shell scripts executable.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and processing data from external cloud provider APIs, creating a surface for indirect prompt injection.
Ingestion points: Data is pulled from AWS Cost Explorer, GCP Billing, and Azure Cost Management APIs as specified in SKILL.md.
Boundary markers: The instructions lack explicit boundary markers or instructions to the agent to ignore potentially malicious content embedded within the billing data or resource tags.
Capability inventory: The skill has access to powerful Bash tools scoped to aws:*, gcloud:*, and az:*, as well as the ability to write files and modify their permissions via scripts/generate_cost_report.py.
Sanitization: There are no defined sanitization or validation steps for the external data before it is interpolated into reports or used to generate infrastructure-as-code (Terraform/CloudFormation) changes.

optimizing-cloud-costs